CN105205500A - Vehicle detecting method based on multi-target tracking and cascade classifier combination - Google Patents

Vehicle detecting method based on multi-target tracking and cascade classifier combination Download PDF

Info

Publication number
CN105205500A
CN105205500A CN201510633007.1A CN201510633007A CN105205500A CN 105205500 A CN105205500 A CN 105205500A CN 201510633007 A CN201510633007 A CN 201510633007A CN 105205500 A CN105205500 A CN 105205500A
Authority
CN
China
Prior art keywords
pix
target tracking
agglomerate
cascade classifier
haar
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201510633007.1A
Other languages
Chinese (zh)
Inventor
黄治同
张雪
纪越峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN201510633007.1A priority Critical patent/CN105205500A/en
Publication of CN105205500A publication Critical patent/CN105205500A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/08—Detecting or categorising vehicles

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)
  • Traffic Control Systems (AREA)
  • Image Processing (AREA)

Abstract

本发明公开了一种基于多目标跟踪与级联分类器融合的车辆检测方法，其具体实施步骤如下：提取样本的15种Haar-like特征并进行筛选，设置分类器训练参数进行分类器训练；进行基于多尺度滤波的新目标检测并进行目标跟踪；本发明提供的车辆检测方法具有检测精确度高，处理速度快的优点。

The invention discloses a vehicle detection method based on multi-target tracking and cascade classifier fusion. The specific implementation steps are as follows: extract 15 kinds of Haar-like features of samples and perform screening, and set classifier training parameters to perform classifier training; Carry out new target detection based on multi-scale filtering and target tracking; the vehicle detection method provided by the invention has the advantages of high detection accuracy and fast processing speed.

Description

基于多目标跟踪与级联分类器融合的车辆检测Vehicle Detection Based on Multi-Target Tracking and Cascaded Classifier Fusion

技术领域 technical field

本发明涉及图像处理和计算机视觉领域，具体说来，涉及一种基于多目标跟踪与级联分类器融合的车辆检测方法。 The invention relates to the fields of image processing and computer vision, in particular to a vehicle detection method based on multi-target tracking and cascade classifier fusion.

背景技术 Background technique

近年来随着我国交通运输事业的蓬勃发展，智能交通系统(ITS)的研究和应用越来越得到重视。准确、实时、完整的交通信息采集是ITS的基础，而车辆检测器则是对动态交通信息进行实时采集的基础设施。随着软硬件技术的飞速发展，各种类型的车辆检测器迅速崛起。主要有感应线圈检测器、磁力检测器、微波检测器、超声波检测器、红外线检测器和视频检测器等，目前我国道路监控系统中，使用最多的是感应线圈车辆检测器、视频车辆检测器和微波车辆检测器3种。 In recent years, with the vigorous development of my country's transportation industry, the research and application of Intelligent Transportation System (ITS) has been paid more and more attention. Accurate, real-time and complete traffic information collection is the basis of ITS, and vehicle detectors are the infrastructure for real-time collection of dynamic traffic information. With the rapid development of software and hardware technology, various types of vehicle detectors have emerged rapidly. There are mainly induction coil detectors, magnetic detectors, microwave detectors, ultrasonic detectors, infrared detectors and video detectors. At present, in my country's road monitoring system, induction coil vehicle detectors, video vehicle detectors and There are 3 types of microwave vehicle detectors.

感应线圈检测器是地埋型检测器，可直接提供车辆出现、车辆通过、车辆计数及车道占有率等交通流信息。一般道路均可设置，主要应用在收费站、互通式立交前后、隧道区段、城市道路、停车场等场合。感应线圈检测器前期投入较少、可靠性高，但维护、重新安装困难，需封闭车道、破坏路面，从长期来看运营成本较高，这也是制约其继续快速发展的一个致命因素。 The induction loop detector is a buried detector that can directly provide traffic flow information such as vehicle presence, vehicle passing, vehicle count and lane occupancy rate. General roads can be set, mainly used in toll stations, before and after interchanges, tunnel sections, urban roads, parking lots and other occasions. Inductive loop detectors require less initial investment and high reliability, but are difficult to maintain and reinstall. They need to close driveways and damage road surfaces. In the long run, the operating costs are relatively high, which is also a fatal factor restricting its continued rapid development.

微波检测器是一种工作在微波频段的雷达探测器，能检测车流量、速度、车道占有率和车型等交通流基本信息的非地埋式检测器，中心频率为10.525GHz，工作方式为主动型。在恶劣气候下性能出色，可全天候工作；可以侧向方式检测多车道；可检测静止的车辆；直接检测速度。但是当道路具有铁质的分隔带时，或路侧有障碍物时检测精度下降；检测器安装条件要求较高，侧向安装时需要后置距离；测速精度低。 The microwave detector is a radar detector working in the microwave frequency band. It is a non-buried detector that can detect the basic information of traffic flow such as traffic flow, speed, lane occupancy rate and vehicle type. The center frequency is 10.525GHz and the working mode is active. type. Excellent performance in harsh weather, can work around the clock; can detect multi-lane sideways; can detect stationary vehicles; directly detect speed. However, when the road has an iron divider, or when there are obstacles on the side of the road, the detection accuracy will drop; the detector installation conditions are relatively high, and a rear distance is required for lateral installation; the speed measurement accuracy is low.

视频车辆检测技术是将视频图像处理和计算机图形识别技术相结合的新型数据采集技术，近年来发展迅速，代表了未来交通流信息检测领域的发展方向。它是用视频摄像机作为传感器，在视频范围内设置虚拟线圈，即检测区，车辆进入检测区时使背景灰度值发生变化，而产生检测信号，通过软件的分析和处理，得到交通量、平均车速、占有率、排队长度等交通参数。还可以利用计算机视觉技术对车辆进行定位、识别和追踪，并对检测对象的交通行为进行分析和判断，最终完成各种交通流数据的采集。视频车辆检测器广泛应用于高速公路和城市道路，目前主要应用在道路条件复杂的地段，如高速公路立交、匝道、隧道，城市道路的交叉路口等。随着视频图像处理和计算机图形识别技术的不断提高、应用领域的扩大以及硬件成本的降低，视频车辆检测器的总体造价随之下降，加之后期运营成本较低，其应用范围将不断扩大。但是目前的视频检测技术中大多是单独使用前景检测技术，使得检测的准确度受环境复杂度的影响很大，而分类器的使用在很大程度上解决了这个问题。 Video vehicle detection technology is a new data acquisition technology that combines video image processing and computer graphics recognition technology. It has developed rapidly in recent years and represents the development direction of the future traffic flow information detection field. It uses a video camera as a sensor, and sets a virtual coil in the video range, that is, the detection area. When the vehicle enters the detection area, the background gray value changes to generate a detection signal. Through the analysis and processing of the software, the traffic volume, average Traffic parameters such as vehicle speed, occupancy rate, queue length, etc. It can also use computer vision technology to locate, identify and track vehicles, analyze and judge the traffic behavior of the detected objects, and finally complete the collection of various traffic flow data. Video vehicle detectors are widely used in highways and urban roads. Currently, they are mainly used in areas with complex road conditions, such as highway interchanges, ramps, tunnels, and intersections of urban roads. With the continuous improvement of video image processing and computer graphics recognition technology, the expansion of application fields and the reduction of hardware costs, the overall cost of video vehicle detectors will decrease, and the later operating costs will be lower, and its application range will continue to expand. However, most of the current video detection technologies use foreground detection technology alone, so that the accuracy of detection is greatly affected by the complexity of the environment, and the use of classifiers solves this problem to a large extent.

发明内容 Contents of the invention

本发明提供了一种基于多目标跟踪与级联分类器融合的视频车辆检测方法，本方法可以实现准确的车辆检测。 The invention provides a video vehicle detection method based on multi-target tracking and cascade classifier fusion, and the method can realize accurate vehicle detection.

为了解决车辆检测准确度的问题，本发明的具体实施步骤为： In order to solve the problem of vehicle detection accuracy, the specific implementation steps of the present invention are:

(1)分类器训练，经过CART决策树筛选的多种Haar-like特征进行训练； (1) Classifier training, trained by various Haar-like features screened by CART decision tree;

(2)基于多尺度滤波的新目标检测； (2) New target detection based on multi-scale filtering;

进一步，步骤(1)中选取合适比例的正负样本进行分类器训练是指：将恰当数量的正样本和负样本从样本库中随机选取出来，然后提取正负样本的15种Haar-like特征，通过CART决策树筛选出鲁棒的特征进行分类器训练，得到一个20级的级联分类器。 Further, selecting an appropriate proportion of positive and negative samples for classifier training in step (1) refers to randomly selecting an appropriate number of positive and negative samples from the sample library, and then extracting 15 Haar-like features of the positive and negative samples , through the CART decision tree to screen out robust features for classifier training, and get a 20-level cascade classifier.

进一步，步骤(2)中基于多尺度滤波的新目标检测是指：将视频当前帧中运动目标检测出，然后通过对目标团块进行多尺度滤波，将过滤后的新目标添加到跟踪器。 Further, the new target detection based on multi-scale filtering in step (2) refers to: detecting the moving target in the current frame of the video, and then adding the filtered new target to the tracker by performing multi-scale filtering on the target blob.

本发明的优点在于采用基于多目标跟踪与级联分类器融合的车辆检测方法，该方法在跟踪到当前帧中所有目标后增加了分类器判定，从而在提高了检测的精确度。另外本方法的算法复杂度比较低，能更好的适应目前计算机视觉系统的应用。 The invention has the advantage of adopting a vehicle detection method based on multi-target tracking and cascading classifier fusion, which adds classifier judgment after tracking all targets in the current frame, thereby improving detection accuracy. In addition, the algorithm complexity of this method is relatively low, which can better adapt to the application of the current computer vision system.

附图说明 Description of drawings

图1为本发明实施例的实施流程示意图； Fig. 1 is the implementation flow schematic diagram of the embodiment of the present invention;

图2为正样本实例图； Figure 2 is an example of a positive sample;

图3为负样本实例图； Figure 3 is a negative sample example diagram ;

图4为需要提取的特征模式图； Fig. 4 is a feature pattern diagram that needs to be extracted;

图5为计算特征示例图； Fig. 5 is an example diagram of calculating features;

图6为RAST(x,y)的定义图 Figure 6 is the definition diagram of RAST(x,y)

图7为旋转45°的矩形正负权值划分示例图 Figure 7 is an example diagram of the positive and negative weight division of a rectangle rotated by 45°

图8为训练分类器第一级示例图； Fig. 8 is a first-level example diagram of the training classifier;

图9为检测结果示例图； Figure 9 is an example diagram of test results;

具体实施方式 Detailed ways

为了更好的说明本发明，以下参照附图和实施例对本发明的具体实施做进一步详细的描述。 In order to better illustrate the present invention, the specific implementation of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

如图1所示，本发明的具体实施步骤为： As shown in Figure 1 , the specific implementation steps of the present invention are:

(1)基于经过CART决策树筛选的多种Haar-like特征进行训练。 (1) Training based on a variety of Haar-like features screened by the CART decision tree.

首先建立正负样本库(如图1,2所示)的描述文件，Postive.vec和Negtive.dat。文件中每一行存储一张图片的检索信息，依次为图像块名称，图像块在原始图像的起始位置坐标(left-top)，图像块的高和宽。 First, establish the description files of the positive and negative sample libraries (as shown in Figures 1 and 2 ), Postive.vec and Negtive.dat. Each line in the file stores the retrieval information of a picture, which is the image block name, the starting position coordinates (left-top) of the image block in the original image, and the height and width of the image block.

描述文件建立完成后，进行的是图像块haar特征提取，特征模式如图4所示，一共包含15种。其中黑色区域的权值为负，白色区域的权值为正。为了提高计算效率，我们使用积分图像来计算图像块的特征值。下面分别介绍计算竖直矩阵型和旋转45°的矩形特征值的计算过程。 After the description file is established, the haar feature extraction of the image block is carried out. The feature modes are shown in Figure 4 , and there are 15 types in total. The weight of the black area is negative, and the weight of the white area is positive. To improve computational efficiency, we use integral images to compute feature values of image patches. The calculation process of calculating the vertical matrix type and the rectangular eigenvalues rotated by 45° are introduced respectively below.

竖直特征值的计算过程：图5为图像的积分图像，则A区域的像素值(记为PixA)，B区域的像素值(记为PixB)。 The calculation process of vertical eigenvalues: Fig. 5 is the integral image of the image, then the pixel value of area A (denoted as Pix A ), and the pixel value of area B (denoted as Pix B ).

PixPix AA == iii 88 ++ iii 44 -- iii 55 -- iii 77 PixPix BB == iii 99 ++ iii 55 -- iii 66 -- iii 88 ff ee aa tt uu rr ee == PixPix AA -- PixPix BB == (( iii 88 -- iii 77 )) ++ (( iii 66 -- iii 55 )) -- (( iii 55 -- iii 44 )) -- (( iii 99 -- iii 88 )) -- -- -- (( 11 ))

旋转45°的矩形特征值的计算过程：图6为RSAT(x,y)的定义图，图7为旋转45°的矩形正负权值划分示例图。 The calculation process of the eigenvalue of the rectangle rotated by 45°: Figure 6 is the definition diagram of RSAT(x,y), and Figure 7 is an example diagram of the positive and negative weight division of the rectangle rotated by 45°.

RR SS AA TT (( xx ,, ythe y )) == ΣΣ xx ,, ≤≤ xx ,, xx ,, ≤≤ xx -- || ythe y -- ythe y ,, || II (( xx ,, ,, ythe y ,, )) RR SS AA TT (( xx ,, ythe y )) == RR SS AA TT (( xx -- 11 ,, ythe y -- 11 )) ++ RR SS AA TT (( xx -- 11 ,, ythe y )) ++ II (( xx ,, ythe y )) -- RR SS AA TT (( xx -- 22 ,, ythe y -- 11 )) -- -- -- (( 22 ))

将提取出加了标签的特征输入到分类器中进行训练，训练过程参数设置：nstage＝20，npos＝1000，nneg＝3000，w＝40,h＝40,其他参数均适用默认值。最终训练出准确率为95.4％的20级分类器，图8为第一级训练过程。 Input the extracted and tagged features into the classifier for training. The parameters of the training process are set: nstage=20, npos=1000, nneg=3000, w=40, h=40, and other parameters are applicable to default values. Finally, a 20-level classifier with an accuracy rate of 95.4% was trained, and Figure 8 shows the first-level training process.

(2)基于多尺度滤波的新目标检测。 (2) New object detection based on multi-scale filtering.

前景检测主要是通过背景建模的方法实现的，这里我们使用的是GMM背景建模。团块检测核心部分是新团块检测：首先从前景图像中检测出所有团块，然后将较小的团块(可能是由噪声引起的)和与已经被跟踪团块有重叠的团块丢弃，并对剩余的团块按照大小顺序排列，只保留其中几个比较大的团块(默认为10)。最后利用多尺度滤波规则筛选，只有在筛选中有返回RECT结果的目标团块才是符合标准的团块，将真正的新团块保存到团块列表中。此时完成新目标检测的任务，将新目标添加到跟踪器，进行后续处理，最终的检测结果如图9所示。 Foreground detection is mainly realized by background modeling method, here we use GMM background modeling. The core part of clump detection is new clump detection: first detect all clumps from the foreground image, and then discard smaller clumps (probably caused by noise) and clumps that overlap with already tracked clumps , and arrange the remaining clumps in order of size, and only keep a few relatively large clumps (the default is 10). Finally, the multi-scale filtering rules are used to screen, and only the target clusters that return RECT results in the screening are clusters that meet the criteria, and the real new clusters are saved in the cluster list . At this point, the task of new target detection is completed, and the new target is added to the tracker for subsequent processing. The final detection result is shown in Figure 9 .

本实施例是在配置为3.60GHzIntel(R)Xeon(R)E5-1620CPU和8G内存的计算机中采用C++编程实现的，每秒处理21帧640*480的图像并且检测准确率达到92.6％。 This embodiment is implemented in a computer configured with 3.60GHz Intel(R) Xeon(R) E5-1620CPU and 8G memory by using C++ programming, processing 21 frames of 640*480 images per second and the detection accuracy rate reaches 92.6%.

Claims (3)

1., based on the vehicle checking method that multiple target tracking and cascade classifier merge, it is characterized in that the concrete steps of the method are:
Step (1), sorter is trained, and the multiple Haar-like feature through the screening of CART decision tree is trained;
Step (2), the fresh target based on multi-scale filtering detects.
2. the vehicle checking method merged based on multiple target tracking and cascade classifier according to right 1, it is characterized in that sorter training in step (1), train based on the multiple Haar-like feature through the screening of CART decision tree, concrete steps are as follows:
S2.1: set up positive and negative pattern representation file.
S2.2: the 15 kinds of haar-like features extracting sample according to formula (1) (2), and carry out sorter training.
Pix A = ii 8 + ii 4 - ii 5 - ii 7 Pix B = ii 9 + ii 5 - ii 6 - ii 8 f e a t u r e = Pix A - Pix B = ( ii 8 - ii 7 ) + ( ii 6 - ii 5 ) - ( ii 5 - ii 4 ) - ( ii 9 - ii 8 ) - - - ( 1 )
Pix in formula aand Pix bbe respectively the pixel value of region A and region B, ii mrepresent the value on integral image block summit, feature is eigenwert.
The pixel that RSAT (x, y) is the 45 ° of regions in region, 45 °, point (x, the y) upper left corner and the lower left corner and, calculate the rectangular characteristic value of rotation 45 ° and calculate the difference being positioned at the capable rectangle RSAT (x, y) of cross exactly.
3. the vehicle checking method merged based on multiple target tracking and cascade classifier according to right 1, it is characterized in that the fresh target based on multi-scale filtering in step (3) detects, concrete steps are as follows:
S3.1: set up background model and carry out foreground detection.
S3.2: carry out new agglomerate detection according to the prospect detected.
S3.3: according to the foreground image in S3.1 and S3.2, carries out multiple dimensioned window filtering by prospect agglomerate, and window smallest dimension is that agglomerate is wide high by 1/2, and be agglomerate size to the maximum, the change of scale of window is 1.1.
S3.4: the agglomerate by filtering is demarcated position fresh target and adds tracker to.
CN201510633007.1A 2015-09-29 2015-09-29 Vehicle detecting method based on multi-target tracking and cascade classifier combination Pending CN105205500A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510633007.1A CN105205500A (en) 2015-09-29 2015-09-29 Vehicle detecting method based on multi-target tracking and cascade classifier combination

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510633007.1A CN105205500A (en) 2015-09-29 2015-09-29 Vehicle detecting method based on multi-target tracking and cascade classifier combination

Publications (1)

Publication Number Publication Date
CN105205500A true CN105205500A (en) 2015-12-30

Family

ID=54953172

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510633007.1A Pending CN105205500A (en) 2015-09-29 2015-09-29 Vehicle detecting method based on multi-target tracking and cascade classifier combination

Country Status (1)

Country Link
CN (1) CN105205500A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106295636A (en) * 2016-07-21 2017-01-04 重庆大学 Passageway for fire apparatus based on multiple features fusion cascade classifier vehicle checking method
CN109697393A (en) * 2017-10-23 2019-04-30 北京京东尚科信息技术有限公司 Person tracking method, device, electronic device and computer-readable medium
CN112558049A (en) * 2019-09-25 2021-03-26 罗伯特·博世有限公司 Method and control device for detecting objects in the surroundings of a vehicle

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104318588A (en) * 2014-11-04 2015-01-28 北京邮电大学 Multi-video-camera target tracking method based on position perception and distinguish appearance model

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104318588A (en) * 2014-11-04 2015-01-28 北京邮电大学 Multi-video-camera target tracking method based on position perception and distinguish appearance model

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
RAINER LIENHART 等: "An Extended Set of Haar-like Features for Rapid Object Detection", 《PROCEEDINGS OF INTERNATIONAL CONFERENCE ON IMAGE PROCESSING》 *
RN DASCHOUDHARY 等: "Real-time Face Detection and Tracking Using Haar Classifier on SoC", 《PROCEEDINGS OF 12TH SARC-IRF INTERNATIONAL CONFERENCE》 *
杨静伟: "基于粒子滤波的行人跟踪", 《中国优秀硕士学位论文全文数据库信息科技辑》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106295636A (en) * 2016-07-21 2017-01-04 重庆大学 Passageway for fire apparatus based on multiple features fusion cascade classifier vehicle checking method
CN109697393A (en) * 2017-10-23 2019-04-30 北京京东尚科信息技术有限公司 Person tracking method, device, electronic device and computer-readable medium
US11270126B2 (en) 2017-10-23 2022-03-08 Beijing Jingdong Shangke Information Technology Co., Ltd. Person tracking method, device, electronic device, and computer readable medium
CN112558049A (en) * 2019-09-25 2021-03-26 罗伯特·博世有限公司 Method and control device for detecting objects in the surroundings of a vehicle

Similar Documents

Publication Publication Date Title
Ghosh On-road vehicle detection in varying weather conditions using faster R-CNN with several region proposal networks
Liu et al. A vision-based pipeline for vehicle counting, speed estimation, and classification
CN103116987B (en) Traffic flow statistic and violation detection method based on surveillance video processing
He et al. Urban rail transit obstacle detection based on Improved R-CNN
CN110619279B (en) Road traffic sign instance segmentation method based on tracking
Yaghoobi Ershadi et al. Robust vehicle detection in different weather conditions: Using MIPM
Ding et al. Fast lane detection based on bird’s eye view and improved random sample consensus algorithm
CN109101939A (en) Determination method, system, terminal and the readable storage medium storing program for executing of state of motion of vehicle
Azimjonov et al. A vision-based real-time traffic flow monitoring system for road intersections
Dinh et al. Development of a tracking-based system for automated traffic data collection for roundabouts
Hu et al. Traffic density recognition based on image global texture feature
Swathy et al. Survey on vehicle detection and tracking techniques in video surveillance
Al Okaishi et al. A vehicular queue length measurement system in real-time based on SSD network
Saha et al. Developing a framework for vehicle detection, tracking and classification in traffic video surveillance
Ribeiro et al. A new microscopic approach to traffic flow classification using a convolutional neural network object detector and a multi-tracker algorithm
Ren et al. Lane Detection in Video‐Based Intelligent Transportation Monitoring via Fast Extracting and Clustering of Vehicle Motion Trajectories
Al Bayati et al. Real-time vehicle detection for surveillance of river dredging areas using convolutional neural networks
CN105205500A (en) Vehicle detecting method based on multi-target tracking and cascade classifier combination
Khalid et al. An android application for unwanted vehicle detection and counting
Gupta et al. YOLO-Based Vehicle Detection and Counting for Traffic Control on Highway
Guerrieri et al. Traffic flow variables estimation: An automated procedure based on moving observer method. Potential application for autonomous vehicles
Ghosh A modified YOLO model for on-road vehicle detection in varying weather conditions
Špaňhel et al. Detection of traffic violations of road users based on convolutional neural networks
CN105069407B (en) A kind of magnitude of traffic flow acquisition methods based on video
Garvanov et al. Detection and parameters estimation of moving objects via video surveillance

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20151230